The Quest in a Generated World

Key insights

6 takeaways

01
Procedural world generation requires quest structures—defined by setting, space, challenge, and goal—to provide meaningful context and player direction.See it on page 3
02
The Charbitat prototype, built on the Unreal Tournament 2004 engine, utilizes a graph-based analysis to integrate key-and-lock puzzles into infinite, procedurally generated 3D environments.See it on page 4
03
World generation relies on an incremental construction method that scores candidate tile configurations using approximately twelve distinct design evaluators.See it on page 6
04
Evaluators enforce structural logic, such as ensuring that locks are placed before their corresponding keys to maintain coherent player progression.See it on page 5
05
Qualitative observations indicate that players demonstrate higher engagement levels in quest-enhanced procedural environments compared to unstructured open spaces.See it on page 6
06
The non-repeating nature of procedurally generated worlds presents a significant challenge for conducting systematic, comparative player research.See it on page 6

Summary

This paper presents a method for generating quests within procedurally generated three-dimensional game worlds, arguing that procedural content creation alone lacks the context necessary to make game spaces meaningful. The authors contend that quests supply this missing structure, and they define a quest through four core elements: setting, space, challenge, and goal. Their prototype, Charbitat, a modification of Unreal Tournament 2004, generates an infinite world as players explore it, partitioning space into tiles and using a graph-based analysis to place key and lock puzzles in a way that structures player progress.

The technical approach combines terrain generation from heightmaps with rule-based population of flora and fauna, and a backend that scores candidate tile configurations using roughly a dozen evaluators. These evaluators encode design rules, such as requiring several locks to appear before a corresponding key, and favor configurations that yield good overall world structure rather than local optima. The authors frame this incremental, rather than holistic, construction as a source of occasional difficulty in making the world cohere.

The findings are presented as twofold: a technical demonstration that a flexible content and context generator can be integrated into an existing engine, and an informal observation that players engaged more with quest-enhanced procedural worlds than with unstructured open spaces. The evidence for the latter is qualitative, drawn from demonstrations and play sessions with visitors and researchers, and the authors acknowledge that the unique, non-repeating nature of each generated world makes systematic player comparison difficult. The paper therefore positions its contribution as a foundation for further work on evaluation frameworks and on personalized quest generation.